Short- and Long-Term Impact of Hyperoxia on the Blood and Retinal Cells’ Transcriptome in a Mouse Model of Oxygen- Induced Retinopathy

www.nature.com/pr

BASIC SCIENCE ARTICLE OPEN Short- and long-term impact of hyperoxia on the blood and retinal cells’ transcriptome in a mouse model of oxygen- induced retinopathy

Magdalena Zasada1, Anna Madetko-Talowska2, Cecilie Revhaug3,4, Anne Gro W. Rognlien3,4, Lars O. Baumbusch3,Teoﬁla Książek2, Katarzyna Szewczyk2, Agnieszka Grabowska2, Miroslaw Bik-Multanowski2, Jacek Józef Pietrzyk1, Przemko Kwinta1 and Ola Didrik Saugstad3,4

BACKGROUND: We aimed to identify global blood and retinal gene expression patterns in murine oxygen-induced retinopathy (OIR), a common model of retinopathy of prematurity, which may allow better understanding of the pathogenesis of this severe ocular prematurity complication and identiﬁcation of potential blood biomarkers. METHODS: A total of 120 C57BL/6J mice were randomly divided into an OIR group, in which 7-day-old pups were maintained in 75% oxygen for 5 days, or a control group. RNA was extracted from the whole-blood mononuclear cells and retinal cells on days 12, 17, and 28. Gene expression in the RNA samples was evaluated with mouse gene expression microarrays. RESULTS: There were 38, 1370 and 111 genes, the expression of which differed between the OIR and control retinas on days 12, 17, and 28, respectively. Gene expression in the blood mononuclear cells was signiﬁcantly altered only on day 17. Deptor and Nol4 genes showed reduced expression both in the blood and retinal cells on day 17.

1234567890();,: CONCLUSION: There are sustained marked changes in the global pattern of gene expression in the OIR mice retinas. An altered expression of Deptor and Nol4 genes in the blood mononuclear cells requires further investigation as they may indicate retinal neovascularization. Pediatric Research (2020) 87:485–493; https://doi.org/10.1038/s41390-019-0598-y

INTRODUCTION oxygen-induced retinopathy (OIR).5 In mice, the retinas are mostly Retinopathy of prematurity (ROP), a proliferative retinal vascular avascular after birth, and retinal vascularization occurs in the early disorder affecting premature infants, is currently a major cause of neonatal period, resembling situation observed in the prematurely potentially preventable blindness in children.1 It is estimated that born infants. 5-day long continuous exposure to high oxygen ROP affects 33.2% of neonates with birth weight < 1500 g.2 ROP levels imitates the biphasic retinal vascular response observed in develops in two phases; phase 1 is initiated by hyperoxia and the young infants with ROP.6 Up till now, exploration of the retinal exacerbated by lack of the essential factors provided in utero, gene expression pattern in the murine OIR model was assessed followed by a destruction of existing vessels and arrest of normal partially. Sato et al.7 evaluated only a selected set of genes related vessel development. Phase 2 is driven by increased metabolic to angiogenesis and inflammation. Subsequent studies investi- activity of the hypoxic retina, which leads to both revascularization gated expression of the entire genome, but at the selected time and pathological neovascularization.3 An excessive abnormal points—Ishikawa et al.8 and Yang et al.9 determined a profile of compensatory vessel growth may result in fibrous scar develop- gene expression in murine retinas soon after end of hyperoxia, ment, and ultimately leads to retinal detachment.4 We believe that whereas Recchia et al.10 additionally verified gene expression in better understanding of the pathomechanisms behind ROP rodent retinas during neovascularization phase. The goal of this coupled with an early and effective method of the risk study was to identify potential ROP blood biomarkers and provide stratification of ROP development would allow to take true a more molecular-based understanding of ROP by comparing the preventive actions and/or targeted treatment. Genome-wide gene expression profile at three time points (vaso-obliteration microarray analysis provides a simultaneous examination of the (P12), maximal neovascularization (P17) and during normalization expression of thousands of genes in a single experiment to create of the retinal damage (P28)) in the blood mononuclear cells and a comprehensive picture of dysregulated genes and pathways. A retinas obtained from the mice with and without OIR. We possibility to use the animal model that mimics key features of the hypothesized that supplemental oxygen as compared with air disease provides invaluable tool for studying ROP underlying induced gene expression changes in the blood and retinal cells, mechanisms not only in the blood, but also in the target tissue which varied depending on the length of time following exposure such as retina. The most widely used model is a murine model of to hyperoxia.

1Department of Paediatrics, Jagiellonian University Medical College, Krakow, Poland; 2Department of Medical Genetics, Jagiellonian University Medical College, Krakow, Poland; 3Department of Paediatric Research, Oslo University Hospital Rikshospitalet, Oslo, Norway and 4University of Oslo, Oslo, Norway Correspondence: Przemko Kwinta ([email protected]) These authors contributed equally: M. Zasada, A. Madetko-Talowska Received: 3 June 2019 Revised: 4 September 2019 Accepted: 6 September 2019 Published online: 2 October 2019

© The Author(s) 2019 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 486 METHODS suspended by manual pipetting, then the samples were placed Murine model of OIR immediately at −80 °C. Total RNA was isolated with the use of The Norwegian Animal Research Authority approved all experi- Mouse RiboPure™-Blood RNA Isolation Kit (Invitrogen; Thermo ment protocols. The animals were cared for and handled in Fisher Scientific) according to the manufacturer’s instructions. RNA accordance with the Norwegian Legislation and Directive. OIR was concentration was measured with the use of NanoDrop Spectro- induced in C57BL/6 mice (received from: Jackson Laboratories, Bar photometer (NanoDrop ND-1000; Thermoscientific, Waltham), and Harbor, ME) according to the protocol described by Smith et al.6 RNA quality was determined by TapeStation 4200 (Agilent, Santa Briefly, mice were housed and bred at 24 °C at a 12:12 light/dark Clara). cycle with access to a standard rodent pellets and water ad libitum. On the postnatal day 7 (P7), all pups within each litter Obtaining retinas and total RNA isolation were earmarked with use of surgical micro-scissors. When a pup Whole retinas free of additional tissues were isolated immedi- was picked up from the cage, it received a number drawn by an ately under a dissecting microscope, and put into vials with assistant from blinded tickets containing numbers from 1 to 8. The RNAlater® Stabilization Solution (Thermo Fisher Scientific), kept procedure was repeated until every pup had its own number for 24 h in + 4°Candstoredina−80 °C freezer until further use. within the litter. The smallest pups (according to their body weight To isolate sufficient total RNA for microarray analysis, two retinas on P7) out of litters of >8, or any abnormal pups (with visible fromthesamemousewerepooled.Retinatissueswerethen morphologic abnormalities) were excluded from the experiment. disrupted with the use of TissueRuptor handheld rotor-stator Each litter with a nursing mother was then randomized either to homogenizer (Qiagen, Hilden, Germany) and total RNA from the hyperoxia (75% oxygen) or normoxia (21% oxygen) conditions retina cells was processed by RNeasy Micro Kit (Qiagen, Hilden, (also by drawing a ticket either with a symbol N for normoxia or H Germany) as described by the manufacturer. RNA concentration for hyperoxia), then transferred in their cage into A-Chambers was measured with the use of NanoDrop Spectrophotometer (BioSpherix Ltd, Parish, NY), where they were kept until day 12 (NanoDrop ND-1000; Thermoscientific, Waltham), and RNA (P12). Seventy-five percent oxygen concentration was achieved quality was determined by TapeStation 4200 (Agilent, Santa with a conventional oxygen mixer (oxygen (AGA AS, Oslo, Norway) Clara). plus room air). Each chamber was continuously monitored for oxygen concentration (O2-monitor, ProOX110 BioSpherix Ltd, Validation of the model Parish, NY), temperature and humidity (Thermometer/Hygrometer Eyes from randomly selected 36 pups both from normoxia and Clas Ohlson, Insjon, Sweden), and CO2 level (ProCO2 P120, hyperoxia groups, 1/3 from each harvesting time point, were used BioSpherix Ltd, Parish, NY). An open water source kept air for the verification of the model, according to the methods humidity at 40–50%. Every day there was a 10–20 min break in described in details elsewhere.3 Briefly, the eye was enucleated oxygen supply for airing the chamber and checking the mice. and fixed with 4% Paraformaldehyde Solution (Boster Biological Nursing dams stayed with their own litters throughout the entire Technology, Pleasanton CA). Then under the dissecting micro- experiment. The cages containing pups with their mothers were scope the retina was isolated, incubated in 0.5% Triton X-100 removed from the hyperoxia/normoxia chambers on day P12. (Sigma, St. Louis, MO) and stained in the lectin solution with Alexa Before removing cages with mice from A-Chambers, one more Fluor 594–I21413, Molecular Probes. Then under a dissecting randomization was performed; namely tickets with symbols, microscope retinas were mounted on a microscope slide and indicating the predicted harvesting day and type of analyses cover slipped with SlowFade medium (Thermo Fisher Scientific, were prepared, then they were randomly assigned to each pup Waltham, MA). Images were obtained using the Zeiss AxioCam from either hyperoxia or normoxia group. 1/3 of the pups were MRc, Zeiss AxioObserver.Z1 microscope and AxioVision 4.8 soft- immediately anesthetized (zolazepam/tiletamine/xylazine/fenta- ware. Using Adobe Photoshop CS6, the percentage of avascular nyl standard ZRF cocktail, 0.1 ml/10 g intraperitoneally), their retina, as well as neovascularization fraction was calculated by an gender was assessed, and the blood samples obtained by cardiac independent researcher. puncture were collected to 1 ml syringes (Terumo U40/1 ml, needle 0.33 x 12), flushed with EDTA, then processed within 1 h, as Microarray analyses described below. Then mice were killed by aorta cut and their eyes One-hundred nanograms of total RNA was used for each single were enucleated. The rest of the litters with nursing mothers experiment with use of SurePrint G3 Mouse Gene Expression returned to room air and stayed under normal conditions till P17, 8x60K Microarrays (Agilent, Santa Clara) to study the mRNA or P28, according to their randomly selected groups. Their blood expression. Microarray gene expression experiment was per- and retinas were collected and processed the same way as formed according to the manufacturer’s protocol (Two-Color described above. Microarray-Based Gene Expression Analysis—Low Input Quick Amp Labeling ver.6.5). The universal mouse reference RNA Murine blood mononuclear cells purification and total RNA (Agilent Technologies, CA) was used as an internal control. After isolation experiment, the microarrays were scanned by SureScan Microarray Blood sample was diluted with equal amount of sterile phosphate- Scanner (Agilent, Santa Clara) and data was extracted using buffered saline (PBS) solution (GE Healthcare Life Sciences). One Feature Extraction Software (Agilent, Santa Clara). Details of the milliliter of Ficoll-Paque (GE Healthcare Life Sciences) was inserted microarray experiment are available on-line (https://www.ncbi. into a 2 ml Eppendorf centrifuge tube. Diluted blood sample was nlm.nih.gov/geo/query/acc.cgi?acc=GSE130400, https://www. layered carefully onto the Ficoll-Paque. Then sample was ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE130347) centrifuged at 400 gav for 30 min at 18 °C. After removing an upper layer, a lymphocyte layer was transferred into a new 2 ml Quantitative real-time PCR (qRT-PCR) verification of gene Eppendorf tube, filled with PBS solution, mixed and afresh expression centrifuged at 100 gav for 10 min at 18 °C. Then the supernatant To verify the microarray gene expression data, ten genes were was discarded carefully not to disrupt the pellet of the peripheral randomly selected for a method validation using qRT-PCR. blood mononuclear cells (PBMC). Subsequently, PBMCs were The genes represented well described and characterized murine washed again with PBS solution (centrifugation repeated under genes chosen from the popular databases, such as Ensembl, the same conditions). Afterwards 800 µl of lysis solution and 50 µl NCBI, or DAVID. The genes specific for retinas included:Cidea, of sodium acetate (Mouse RiboPure™-Blood RNA Isolation Kit; Etnppi, Spink13, Gabrb1, Bfsf2, Tnnt2, Adm, Gfap, Nupr, Actg2. Invitrogen; Thermo Fisher Scientific) was added, the cells were The primers and probes were commercially available (Thermo

Pediatric Research (2020) 87:485 – 493 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 487 Fischer Scientific, USA). The complementary DNA was synthe- RESULTS sized using SuperScript III kit(ThermoFischerScientific) Animal model of OIR according to the manufacturer’s protocol. The qRT-PCR analysis On P7, 120 C57BL/6J mice underwent the experiment—60 were was performed using Applied Biosystems 7500 Real-Time PCR randomly assigned to the OIR group, and the other 60 formed the System and TaqMan Universal Master Mix II, no UNG (Thermo control group. All animals survived the experiment, and were Fischer Scientific). The qRT-PCR reactions were performed in 96- further randomized according to the day of tissue harvesting (P12, well plates at 95 °C for 10 min, followed by 40 cycles of 95 °C for P17, or P28). Six mice from each group from every experimental 15 s and 62 °C for 1 min. All samples were run in duplicates. time point were used for model verification. Out of 84 animals Gene expression was calculated by ΔΔCt method, normalized assigned to the gene expression analysis, five were excluded due against the endogenous control gene Gapdh and for each gene to insufficient RNA concentration and low sample quality (Fig. 1). fold change (FC) values were calculated as previously The two groups did not differ in gender distribution (Table 1). described.11 In blood samples, the same validation method as There were no significant weight differences between the groups described above was performed in seven randomly selected when entering experiment (P7); however, OIR group revealed genes, due to very limited amount of blood and RNA obtained transient significant extrauterine growth retardation on P12, which from the small pups. normalized after day P17 (Table 2). Retinal vascular staining confirmed oxygen-related damage of the retinas in the OIR group Statistical analysis (Fig. 2). Gender distribution was evaluated with chi-square test. A probability value of < 0.05 was considered statistically significant. Overall gene expression changes in retina JMP® 13.1.0 (SAS Institute Inc., 2016) was used for statistical Microarray analysis revealed significant differences between OIR analysis. and control group in gene expression in murine retinas in all Microarray data were analyzed by GeneSpring GX software experimental time points (Table 3 and Fig. 3—part 1). Overall, (Agilent, Santa Clara). Raw microarray data were normalized, then there were 39 genes (38 genes with annotated gene symbol) moderated t-test was applied to compare gene expression differentially expressed on day P12, 1478 genes (1370 genes with between the study groups. The results were corrected for multiple annotated gene symbol) on P17, and 114 genes (111 genes with testing by the Benjamini–Hochberg procedure. A 0.05 significance annotated gene symbol) on P28. On P12, the predominant level of the adjusted p-values was used. The genes, the expression fraction of the significantly altered genes showed reduced of which was significantly altered during the experiment, were expression, whereas the majority of the genes demonstrated further investigated using established databases, such as Gene increased expression in later days, both P17 and P28, in the OIR Cards (http://www.genecards.org), Mouse genome informatics group when compared with the controls. We observed a shift in (http://www.informatics.jax.org), and NCBI gene (http://www.ncbi. gene expression pattern in OIR murine retinas, starting from a nlm.nih.gov), Ensembl (http://www.ensembl.org/Mus_musculus). significantly diminished gene expression immediately after

C57BL/6J mice (n = 120)

Postnatal day 7

Hyperoxia Normoxia

FiO2* 0.75, 5 days FiO2* 0.21, 5 days (n = 60) (n = 60)

Postnatal day 12 Postnatal day 12 Harvested for blood/retinas samples (n = 14) Harvested for blood/retinas samples (n = 14) Included into gene analyses = 13 Included into gene analyses = 13 Excluded from analyses = 1 Excluded from analyses = 1 Model verification (n = 6) Model verification (n = 6)

Postnatal day 17 Postnatal day 17 Harvested for blood/retinas samples (n = 14) Harvested for blood/retinas samples (n = 14) Included into gene analyses = 13 Included into gene analyses = 13 Excluded from analyses = 1 Excluded from analyses = 1 Model verification (n = 6) Model verification (n = 6)

Postnatal day 28 Postnatal day 28 Harvested for blood/retinas samples (n = 14) Harvested for blood/retinas samples (n = 14) Included into gene analyses = 13 Included into gene analyses = 14 Excluded from analyses = 1 Excluded from analyses = 0 Model verification (n = 6) Model verification (n = 6)

Fig. 1 Flowchart of the study. Experimental design and experimental groups. *FiO2—fraction of inspired oxygen

Pediatric Research (2020) 87:485 – 493 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 488

Table 1. Characteristics of dams and litters in the study groups

Hyperoxia Normoxia Pearson chi-square p— for gender distribution No of dams/litters (n) Male pups (n) Female pups (n) No of dams/litters (n) Male pups (n) Female pups (n)

In total 18 20 19 16 21 19 0.5400 P12 12 5 8 12 8 5 0.2393 P17 10 10 3 10 7 6 0.2162 P28 10 5 8 9 6 8 0.8163

hyperoxia exposure through a peak gene overexpression on P17, when maximal neovascularization was observed, till gradual Table 2. Mean weight in selected time points in OIR and control decline of gene overexpression on P28 (Fig. 3). animals, whose tissues were further proceeded for the transcriptome The genes with the most diverse expression defined by the analysis highest fold change (FC) values are listed in Tables 4 and 5. OIR group Normoxia group p (two sided t-test) Overall gene expression changes in blood Weight on postnatal day (P) [g]: There were significant differences in gene expression only on P17 P7 (mean, (SD)) 4.13 (0.41) 3.98 (0.59) 0.2092 in OIR mice in comparison to the control animals in the PBMC fi P10 (mean, (SD)) 5.31 (0.54) 5.29 (0.53) 0.8675 analysis (Fig. 3). Out of 63 genes (49 genes with known of cial Gene symbols), around two-thirds were significantly upregulated, P12 (mean, (SD)) 5.58 (0.54) 6.06 (0.56) 0.0003 whereas one-third was downregulated (Table 3 and Fig. 3—part P14 (mean, (SD)) 6.13 (0.68) 6.92 (0.80) 0.0004 2). Two genes, Deptor and Nol4 showed decreased expression P17 (mean, (SD)) 7.03 (0.78) 7.53 (0.84) 0.0268 both in the blood and retinal tissues on day P17 (FCretina < −3; P21 (mean, (SD)) 8.49 (1.19) 8.86 (1.02) 0.3939 FCblood < −2). P28 (mean, (SD)) 14.96 (1.93) 16.07 (1.34) 0.1286 Quantitative real-time PCR The validation procedure performed for genes randomly selected before experiments did not reveal significant differences between the results obtained with use of microarrays as compared with real-time PCR method (t-test p-values for each time point > 0.05). Detailed results for blood and retina tissue are shown in Fig. 4. P12H P17H P28H

acf DISCUSSION The goal of the study was to identify short- and long-term molecular alterations in the blood and retina after exposure to high levels of inhaled oxygen using a well-characterized murine model of oxygen-induced retinopathy. The nature of the study was mainly exploratory. On P12, the genes with the most decreased expression were bdg those associated with angiogenesis (Ccm2l, Mmrn2, Ets1, Ptprb, CD34, Robo4), cell shape rearrangement (Arap3), ion channel formation (Abcc9), as well as receptors involved in cell-to-cell interactions (Adgre5). Interestingly, some of the genes regulating angiogenesis were first underexpressed on P12, then they were overexpressed on P17 to finally normalize their expression on P28. An example of such gene response, the expression of which e reversed between P12 and P17, was Cd34 gene. Cd34 encodes a sialomucin CD34, ubiquitously expressed in the luminal membrane of the endothelial cells, which promotes angiogenesis.12 CD34 transcript and protein levels are increased during human angiogenesis.13 According to Siemerink et al.13 its expression promotes pathological neovascularization in a murine OIR model by enhancing epi-retinal tuft formation. Fig. 2 Retinal flat mounts from OIR groups at different time points: On P17, the most profound decrease in expression was P12H (postnatal day 12, right after hyperoxia), P17H (postnatal day observed in the genes associated with: apoptosis (Cidea), 17, 5 days after hyperoxia), P28H (postnatal day 28, 16 days after metabolism (Etnppl), cell division (Spc25), ion channels (Kcne1l, hyperoxia). The retinal vessels were stained with lectin. P12H: a vaso- Kcnd2, Gabrb1, Scn11a), and receptor components (Garb1, Tacr3, obliteration showed as a large avascular zone in the central retina, b Gpr101), as well as serine protease inhibitors (Spink13). the avascular area presented in white; P17H: c maximal pathological On the same day, P17, the top overexpressed genes were Esm1 retinal neovascularization (NV). The retinal vessels were dilated and fi there were tufts of NV on the surface of the retina, d the avascular and Edn2 genes. Esm1 encodes endothelial cell-speci c molecule 1 area shown in white, e the areas of neovascularization shown in (Esm1). Animal model showed that Esm1, in cooperation with white; P28H: f neovascular changes regression, g the remained areas vascular endothelial growth factor (VEGF), plays an important role of neovascularization are shown in white in both physiologic and pathological angiogenesis and vascular

Pediatric Research (2020) 87:485 – 493 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 489

Table 3. Summary of the number of differentially expressed genes in retina tissue and blood between OIR and control group at different time points (cutoff: moderated t-test corrected p-value < 0.05)

No of differentially expressed genes with annotated gene symbol P12 P17 P28 (no of differentially expressed microarray probes) Retina Blood Retina Blood Retina Blood

All 38 (39) 0 (0) 1370 (1478) 49 (63) 111 (114) 0 (0) Overexpressed genes 1 0 855 34 (38) 107 0 Underexpressed genes 37 0 515 15 (25) 4 0

I. abc 4 4 4

3 3 3 -value) -value) -value) P P P

2 2 2

1 1 1 –log10 (corrected –log10 (corrected –log10 (corrected

–3 –2 –1 0 1 2 3 –6 –4 –2 0 2 4 6 –3 –2 –1 0 1 2 3 log2 (fold change) log2 (fold change) log2 (fold change)

II. ab c 3 1 0.4 2.5 0.8 -value) -value) -value) 0.3 P P P 2 0.6 1.5 0.2 0.4 1 0.1 –log10 (corrected –log10 (corrected –log10 (corrected 0.2 0.5

–3 –2 –1 0 1 2 3 –4–2–101234–3 –2 –1 0 1 2 log2 (fold change) log2 (fold change) log2 (fold change) Fig. 3 Volcano plots demonstrating that OIR was associated with alterations of specific genes in all analyzed time points (a—P12, b —P17, c—P28). I—gene expression changes in retina, II—gene expression changes in blood. The vertical axis value—negative log of the p-values. The horizontal axis—log2 change (FC—fold change) between the two experimental conditions (OIR vs. normoxia). The red dots represent genes, whose expression was significantly changed between the study vs. control groups (corrected p-value < 0.05) permeability. It is also involved in some inflammatory processes by In previous studies, murine retinal neovascularization as regulation of leukocytes extravasation.14 visualized in the tissue samples spontaneously regressed between Edn2 gene, which showed 18.6-fold increased expression on P17 and P25.3 However, our results indicated that despite almost P17, encodes endothelin (ET) 2, a potent vasoconstrictor that complete disappearance of the visible neovascular changes, the contributes to regulation of the retinal blood flow, and may exert alterations in gene expression in the retinal cells remained. In our mitogenic, pro-oxidative, and proinflammatory effects.15 Our study, we also examined gene expression in the retinas from mice results of Edn2 overexpression confirmed previously published on P28, which in mouse terms was considered a "young" adult or findings of increased abundance of Edn2 transcripts in various in human terms, at least an adolescent. We observed sustained retinal injuries.16 A pharmacological blockade of the ET system overexpression of the several important genes (Apln, Pcolce, prevented pathological neovascularization in a murine model of Cebpd, Bcl3, Tnnt2, C4b) from P17 to P28. Apln gene encodes a OIR.17 Other genes that showed increased expression in our study peptide apelin that stimulated angiogenesis and induces retinal were also evaluated elsewhere. Inhibition of adrenomedullin, a neovascularization in the VEGF-independent manner.20 It was peptide involved in vasodilation, angiogenesis and vascular hypothesized, that apelin, after binding to its APJ receptor, leakage, encoded by Adm gene, highly overexpressed in our promotes nitric oxide production through PI3K⁄Akt signaling and study, was identified as a potential target of new therapies in nitric oxide synthase activation.21 Apelin production increases in diabetic retinopathy.18 On the other hand, an increased expres- response to hypoxia and transcription enhancement by hypoxia- sion of Gfap gene might indicate reactive gliosis as a reaction to inducible factor-1.22 Of note, apelin was also found in vitreous retinal impairment.19 humor of the patients with diabetic retinopathy.23 Interestingly,

Pediatric Research (2020) 87:485 – 493 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 490

Table 4. Top list of downregulated genes in retina and blood based on the fold change in their expression between OIR and control mice at different time points

Time point Gene symbol Log2FC Corrected p-value Gene name

Retina P12 Ccm2l −4.15 0.0374 Cerebral cavernous malformations 2 protein-like Abcc9 −3.93 2.0026E-6 ATP-binding cassette subfamily C member 9 Adcy4 −3.10 0.0095 Adenylate cyclase type 4 Arap3 −2.90 0.0075 Arf-GAP with Rho-GAP domain, ANK repeat and PH domain-containing protein 3 Mmrn2 −2.85 0.0230 Multimerin-2 Robo4 −2.82 0.0433 Roundabout homolog 4 Cd34 −2.81 0.0400 Hematopoietic progenitor cell antigen CD34 Adgre5 −2.76 2.6282E-9 Adhesion G-protein-coupled receptor E5 Ptprb −2.69 2.0025E-6 Receptor-type tyrosine-protein phosphatase beta Ets1 −2.66 0.0400 Protein C-ets-1 P17 Cidea −4.44 0.0043 Cell death activator CIDE-A Kcne1l −4.17 0.0169 Potassium voltage-gated channel subfamily E regulatory beta subunit 5 Kcnd2 −3.82 0.0430 Potassium voltage-gated channel subfamily D member 2 Spink13 −3.78 0.0166 Serine protease inhibitor Kazal-type 13 Etnppl −3.69 0.0371 Ethanolamine-phosphate phospho-lyase Spc25 −3.64 7.0628E-9 Kinetochore protein Spc25 Gabrb1 −3.63 0.0163 Gamma-aminobutyric acid receptor subunit beta-1 Tacr3 −3.33 0.0477 Neuromedin-K receptor Scn11a −3.31 0.0350 Sodium channel protein type 11 subunit alpha Gpr101 −3.30 0.0039 Probable G-protein-coupled receptor 101 P28 Fam219aos −2.62 0.0099 Family with sequence similarity 219, member A, opposite strand Grip1 −1.95 0.0405 Glutamate receptor-interacting protein 1 Crabp1 −1.33 0.0417 Cellular retinoic acid-binding protein 1 Blood P17 Gm1043 −5.02 0.0481 Predicted gene 1043 Gm29260 −4.21 0.0298 Predicted gene 29260 Tbc1d17 −3.63 0.0481 TBC1 domain family member 17 Nol4 −3.23 0.0146 Nucleolar protein 4 Gm4013 −3.13 0.0304 Predicted gene 4013 Gm3970 −2.82 0.0102 Predicted gene 3970 Ccl20 −2.76 0.0434 C-C motif chemokine 20 Cers4 −2.34 0.0253 Ceramide synthase 4 Nbeal1 −2.08 0.0115 Neurobeachin-like 1 Serpina11 −2.03 0.0282 Serpin A11

apelin knockout mice treated with VEGF showed impaired with the previously reported studies highlighting the role of angiogenesis,24 therefore we believe that development of apoptosis of the endothelial cells in pathological vascularization apelin/APJ inhibitors might represent a new therapeutic strategy regression.30 Other genes highly overexpressed on P28 were these to prevent pathological neovascularization in ROP. Tnnt2 encodes engaged in the inflammatory processes: Glycam1 encodes a a troponin T type 2, which might control excessive angiogenesis in proteoglycan ligand for L-selectin, which in turn exerts an anti- hypoxic conditions.25 Pcolce gene encodes procollagen C- inflammatory effect by prevention of monocyte extravasation,31 endopeptidase enhancer protein that is required for endothelial Lyz1 encodes lysozyme, a potent antimicrobial protein. The cell lumen formation.26 Cebpd gene encodes a transcription factor basement membrane anchoring filament protein Lad1, encoded CCAAT/enhancer-binding protein D, associated with immune by Lad1 gene, contributes to the epithelial stability.32 Lad1 gene response27, as well as angiogenesis.28 C4b encodes complement expression is regulated by both lipopolysaccharide stimulation C4-B involved in a variety of immune processes. B-cell leukemia/ and glucocorticoid receptor activation.33 Overexpression of the lymphoma 3 (Bcl3) gene serves as death promoter.29 We genes associated with various aspects of immune response might hypothesized that its increased expression on P17 corresponded indicate a local inflammatory process, which continue in the to massive apoptosis of damaged retinal cells. Additionally, on damaged retina over time. Moreover, we observed an increased P28, observed induction of genes involved in apoptosis might expression of Fgf2os gene, encoding fibroblast growth factor 2 result in apoptosis of the endothelial cells, causing the already opposite strand, which was hypothesized to aid in preservation of formed neo-vessels to regress. This observation was in accordance the neuroretinal function during ischemic retinopathy.34

Pediatric Research (2020) 87:485 – 493 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 491

Table 5. Top list of upregulated genes in retina and blood based on their expression fold change between OIR and control mice at different time points

Time point Gene symbol Log2FC Corrected p-value Gene name

Retina P12 Mri1 1.29 0.0338 Methylthioribose-1-phosphate isomerase P17 Esm1 49.02 1.3345E-11 Endothelial cell-specific molecule 1 Edn2 18.60 1.8853E-5 Endothelin-2 Bcl3 16.71 1.4607E-8 B-cell lymphoma 3 protein homolog Apln 16.05 3.7819E-9 Apelin C4b 11.10 1.3346E-11 Complement C4-B Pcolce 10.65 6.6323E-11 Procollagen C-endopeptidase enhancer 1 Cebpd 10.47 1.9108E-8 CCAAT/enhancer-binding protein delta Adm 9.68 7.5001E-11 Adrenomedullin Tnnt2 8.46 3.2014E-8 Troponin T, cardiac muscle Gfap 8.06 7.0122E-6 Glial fibrillary acidic protein P28 Bcl3 4.62 1.2942E-4 B-cell lymphoma 3 protein homolog C4b 4.17 2.9285E-7 Complement C4-B Lad1 3.92 0.0023 Ladinin-1 Apln 3.92 0.0012 Apelin Pcolce 3.64 2.0317E-4 Procollagen C-endopeptidase enhancer 1 Fgf2os 3.22 0.0091 Uncharacterized protein Tnnt2 2.86 2.8065E-5 Troponin T, cardiac muscle Cebpd 2.77 6.8232E-4 CCAAT/enhancer-binding protein delta Lyz1 2.64 0.0099 Lysozyme C-1 Glycam1 2.61 0.0187 Glycosylation-dependent cell adhesion molecule 1 Blood P17 Taf4 3.11 0.0403 Transcription initiation factor TFIID Sec22a 2.96 0.0195 Vesicle-trafficking protein SEC22a Trmt1l 2.91 0.0195 tRNA methyltransferase 1 -like protein ND2 2.50 0.0070 NADH-ubiquinone oxidoreductase chain 2 Rrp1b 2.32 0.0067 Ribosomal RNA processing protein 1 homolog B Trp53 2.25 6.9234E-4 Cellular tumor antigen p53 Fam20b 2.23 0.0481 Glycosaminoglycan xylosylkinase Hmg20b 1.98 0.0067 SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily E member 1-related Gbf1 1.66 0.0304 Golgi-specific brefeldin A-resistance factor-1 Mkln1 1.60 0.0354 Muskelin

Microarray analysis of the blood mononuclear cells revealed The genes with the most decreased expression in the PBMC alterations in gene expression pattern only on P17. Our results included the ones encoding proteins involved in various cellular indicated that examination of the PBMC’s genome could not mechanisms (Serpina11, Tbc1d17,Cers4), immune response (Ccl20), provide a good early biomarker of OIR, which further confirmed as well as genes encoding long non-coding RNA (Gm29260, our findings in the previous human study on ROP.11 Lack of the Gm4013). altered gene expression on P28 indicated transient nature of the There were two genes that showed significantly reduced gene expression changes in the blood MNCs caused by hyperoxia. expression with FC < −2.0 both in the retinal and blood cells, on Moreover, alterations in gene expression present at retina tissue P17. These were Deptor and Nol1 genes. Deptor gene encodes DEP may appear in PBMCs with a shift of time, therefore they are not Domain-Containing mTOR-Interacting Protein, which is a negative visible at the studied time points. It is also possible, that the regulator of mammalian target of rapamycin (mTOR) signaling applied experimental procedure that caused only subtle changes pathways. The mTOR-signaling pathways were previously shown in gene expression at P12 and P28 in the target tissue (retina) was to be profoundly interrelated with several mechanisms crucial for not potent enough to trigger changes visible in the blood cells. It retinopathy progression, including oxidative stress, inflammation, should also be taken into consideration that when performing a hypoxia, angiogenesis, and proliferation/fibrosis,35,36 Therefore, statistical analysis with an amendment to multiple comparisons (in decreased Deptor gene expression in both blood mononuclear the presented experiments Bonferroni correction was applied), cells and retina might not only mark encountered oxidative stress, small differences in gene expression are overlooked. Finally, it is but also indicate effects of mTOR-signaling pathway overactiva- worth to mention that many of the observed gene expression tion, including neovascularization. changes may be tissue-specific. Consequently we may assume, Nucleolar protein 4 (Nol4) gene might be involved in a process that not all changes in expression present in the retina are of determining cell specificity.37 As a tumor suppressor gene reflected in PBMCs. showed tissue-specific distribution,37 it was considered as a

Pediatric Research (2020) 87:485 – 493 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 492 Part 1 Part 2 P12 P12

2 2 1.45 1.28 1.26 1.37 1.35 1.1 1.12 1.29 1.29 1.5 1.01 1.23 1.16 1.15 1.1 1.12 1.11 1 1.06 1 0.5 0 0 –1 –0.5 –2 –1 –1.04 –1.02 –1.08 –1.13 –1.5 –3 –1.26 –1.97 –2.02 –1.01 –1.04 –1.04 –2 –1.07 –1.18

–1.35 –4 –2.5 –1.53 –3.24 –1.73 ETNPPI SPINK13 BFSF2 ADM GFAP NUPR ACTG2 CIDEA ETNPPI SPINK13 GABRB1 BFSF2 TNNT2 ADM GFAP NUPR ACTG2–2.11

FC microarray FC real time PCR FC microarray FC real time PCR

P17 P17 30 2.5 25.83 2.00

25 2 1.75 1.54 20.71 1.26 1.27 1.5 1.24 1.08

20 1.04 1 15 13.89 12.00

9.69 0.5 8.41 8.47

10 7.45 6.93 0 5 8.14 –0.5 0 –1

–5 –1.5 –1 –1.04 –1.06 –2 –1.05 –3.7 –10 –1.25 –2.78 –4.4 –3.25 –3.26 –3.58 –3.63 –3.68 –1.39 –3.79 –4.44 CIDEA ETNPPI SPINK13 GABRB1 BFSF2 TNNT2 ADM GFAP NUPR ACTG2 ETNPPI SPINK13 BFSF2 ADM GFAP NUPR ACTG2 FC microarray FC real time PCR FC microarray FC real time PCR

P28 P28 5 4.39 2 1.48 4 1.5 1.43 1.36 1.1 1.03 3.06 1.02 2.87 2.82

2.58 1 3 2.51 2.07

1.95 0.5 1.59 2 1.53 1.38 1 1.06 1.04 1.05 0 1 –0.5 0 –1 –1 –1.5 –1.05 –1.06 –1.13 –1.18 –2 –2 –1.25 –1.23 –1.15 –1.55 –1.14

–1.24 ETNPPI SPINK13 BFSF2 ADM GFAP NUPR ACTG2 –1.35 –1.36 CIDEA–1.34 ETNPPI SPINK13 GABRB1 BFSF2 TNNT2 ADM GFAP NUPR ACTG2

FC microarray FC real time PCR FC microarray FC real time PCR Fig. 4 Validation of microarray study by real-time PCR technique in all analyzed time points. Part 1—gene expression in retina tissue, part 2— gene expression in blood

potential biomarker of head and neck cancer, as it was highly which consisted of a several types of cells. Our study provided a hypermethylated in the patients suffering from these starting point for further genetic analysis of ROP, in which we malignancies,38,39 Its role in the OIR would need to be elucidated. should investigate regulatory mechanisms of gene expression The genes overexpressed on P17 in the blood were different such as DNA methylation and/or protein synthesis. In addition our from these upregulated in the retinas at the same time point, and analysis was based predominantly on the molecular studies, we mostly they were involved in transcription and its regulation and believe that future experiments should include functional data, double-strand DNA break repair (Aff3, Baz1b,Dlx5, Hmg20b, Lipin2, both in vitro and/or in vivo provided in the specific cell types or in Nelfe, Trp53, Zfp426). There was no surprise to find overexpression whole retina. of Trp53 gene since that was an indicator of DNA damage caused by hyperoxia.40 CONCLUSIONS Limitations Our mRNA expression analysis showed that particular OIR stages Although our experiment included a sufficient number of animals in mice were associated with specific changes in the pattern of and the gene expression analysis was validated and confirmed, we gene expression in the retinas and allowed for selection of the are aware of some study limitations. The model of continuous candidate genes that may be the subject of further, more targeted hyperoxia exposure for 5 days is more similar to the former cases studies (i.e., proteomics). Moreover, it also revealed, that of ROP associated with treatment of neonates with very high transcriptome blood analysis might not be an early biomarker of oxygen levels. Future studies with a model more closely OIR development, although the altered expression of Deptor and mimicking the clinical situation in preterm infants, such as a rat Nol4 gene would require further investigation as they might model of oxygen-induced retinopathy, are required to confirm the become blood indicators of retinal neovascularization. Our work obtained results. Moreover, we measured only the ambient provided new evidence of the critical role of excessive tissue oxygen concentration within the chambers. We did not check oxygenation that resulted in gene expression changes during the SpO2 or pO2 in the mice subjected to hyperoxia during the course of each stage of retinopathy of prematurity. The altered hyperoxic period (in order to more precisely monitor hyperoxic gene expression in retinas might present a long-lasting effect of a exposure). Additionally, we observed transient extrauterine prior hyperoxia, the subsequent consequences of which would growth retardation in our OIR mice, which might have affected have to be determined in the future studies. their phenotype. However, we did not observe mice weighing <6 g on P17, that according to Connor et al.3 might have profoundly altered the vascular phenotype. Validation of gene expression ACKNOWLEDGEMENTS changes was performed only for selected genes. We understand We thank Monika Atneossen-Asseng and Grethe Dyrhaug for technical assistance and that we presented gene expression results for the entire retina, Agate Noer and Wojciech Zasada for cooperation during obtaining retina pictures.

Pediatric Research (2020) 87:485 – 493 Short- and long-term impact of hyperoxia on the blood and retinal cells’... M Zasada et al. 493 This work was supported by Polish-Norwegian Research Program, operated by the 21. Kojima, Y. & Quertermous, T. Apelin-APJ signaling in retinal angiogenesis. Arter- National Center for Research and Development under the Norwegian Financial ioscler. Thromb. Vasc. Biol. 28, 1687–1688 (2008). mechanism 2009–2014 in the frame of Project Contract No Pol-Nor/196065/54/2013. 22. Eyries, M. et al. Hypoxia-induced apelin expression regulates endothelial cell proliferation and regenerative angiogenesis. Circ. Res. 103, 432–440 (2008). 23. Tao, Y. et al. Apelin in plasma and vitreous and in fibrovascular retinal mem- AUTHOR CONTRIBUTIONS branes of patients with proliferative diabetic retinopathy. Invest. Ophthalmol. Vis. – Substantial contributions to conception and design—M.Z., C.R., A.M.T., A.G.W.R., O.D.S., Sci. 51, 4237 4242 (2010). fi M.B.M., J.J.P., P.K.; acquisition of data—M.Z., A.M.T., C.R., A.G.W.R., T.K., K.S., A.G.; analysis 24. Kasai, A. et al. Retardation of retinal vascular development in apelin-de cient – and interpretation of data—A.M.T., M.Z.; drafting the article or revising it critically for mice. Arterioscler. Thromb. Vasc. Biol. 28, 1717 1722 (2008). fl important intellectual content—M.Z., A.M.T., C.R., L.O..B, O.D.S., M.B.M., J.J.P., P.K.; final 25. Watson, O. et al. Blood ow suppresses vascular Notch signalling via dll4 and is approval of the version to be published—all authors. required for angiogenesis in response to hypoxic signalling. Cardiovasc. Res. 100, 252–261 (2013). 26. Newman, A. C. et al. The requirement for fibroblasts in angiogenesis: fibroblast- ADDITIONAL INFORMATION derived matrix proteins are essential for endothelial cell lumen formation. Mol. Biol. Cell. 22, 3791–3800 (2011). Competing interests: The authors declare no competing interests. 27. Corbi, S. C. T. et al. Expression profile of genes potentially associated with ade- quate glycemic control in patients with type 2 diabetes mellitus. J. Diabetes Res. ’ Publisher s note Springer Nature remains neutral with regard to jurisdictional claims 2017, 2180819 (2017). fi in published maps and institutional af liations. 28. Ko, C. Y. et al. CCAAT/enhancer-binding protein delta/miR135a/thrombospondin 1 axis mediates PGE2-induced angiogenesis in Alzheimer's disease. Neurobiol. Aging 36, 1356 68 (2015). REFERENCES 29. Natoli, R., Provis, J., Valter, K. & Stone, J. Gene regulation induced in the C57BL/6J 1. Blencowe, H. et al. Preterm-associated visual impairment and estimates of reti- mouse retina by hyperoxia: a temporal microarray study. Mol. Vis. 14, 1983–1994 nopathy of prematurity at regional and global levels for 2010. Pediatr. Res. Suppl (2008). 1,35–49 (2013). 30. Volpert, O. V. et al. Inducer-stimulated Fas targets activated endothelium for 2. Cavallaro, G. et al. The pathophysiology of retinopathy of prematurity: an update destruction by anti-angiogenic thrombospondin-1 and pigment epithelium- of previous and recent knowledge. Acta Ophthalmol. 92,2–20 (2014). derived factor. Nat. Med. 8, 349–357 (2002). 3. Connor, K. M. et al. Quantification of oxygen-induced retinopathy in the mouse: a 31. Williams, P. A. et al. GlyCAM1 negatively regulates monocyte entry into the optic model of vessel loss, vessel regrowth and pathological angiogenesis. Nat. Protoc. nerve head and contributes to radiation-based protection in glaucoma. J. Neu- 4, 1565–1573 (2009). roinflammation. 14, 93 (2017). 4. Sapieha, P. et al. Retinopathy of prematurity: understanding ischemic retinal 32. Motoki, K., Megahed, M., LaForgia, S. & Uitto, J. Cloning and chromosomal vasculopathies at an extreme of life. J. Clin. Invest. 120, 3022–3032 (2010). mapping of mouse ladinin, a novel basement membrane zone component. 5. Hartnett, M. E. & Penn, J. S. Mechanisms and management of retinopathy of Genomics 39, 323–330 (1997). prematurity. N. Engl. J. Med. 367, 2515–2526 (2012). 33. Bixler, G. V. et al. Chronic insulin treatment of diabetes does not fully normalize 6. Smith, L. E. et al. Oxygen-induced retinopathy in the mouse. Invest. Ophthalmol. alterations in the retinal transcriptome. BMC Med. Genomics. 4, 40 (2011). Vis. Sci. 35, 101–111 (1994). 34. Fang, L., Barber, A. J. & Shenberger, J. S. Regulation of fibroblast growth factor 2 7. Sato, T. et al. Comprehensive gene-expression profile in murine oxygen-induced expression in oxygen-induced retinopathy. Invest. Ophthalmol. Vis. Sci. 56, retinopathy. Br. J. Ophthalmol. 93,96–103 (2009). 207–215 (2014). 8. Ishikawa, K. et al. Gene expression profile of hyperoxic and hypoxic retinas in a 35. Sasore, T., Reynolds, A. L. & Kennedy, B. N. Targeting the PI3K/Akt/mTOR pathway mouse model of oxygen-induced retinopathy. Invest. Ophthalmol. Vis. Sci. 51, in ocular neovascularization. Adv. Exp. Med. Biol. 801, 805–811 (2014). 4307–4319 (2010). 36. Nakahara, T. et al. Mammalian target of rapamycin (mTOR) as a potential ther- 9. Yang, X., Dong, X., Jia, C. & Wang, Y. Profiling of genes associated with the murine apeutic target in pathological ocular angiogenesis. Biol. Pharm. Bull. 40, model of oxygen-induced retinopathy. Mol. Vis. 19, 775–788 (2013). 2045–2049 (2017). 10. Recchia, F. M. et al. Identification of genes and pathways involved in retinal 37. Ueki, N. et al. NOLP: identification of a novel human nucleolar protein and neovascularization by microarray analysis of two animal models of retinal determination of sequence requirements for its nucleolar localization. Biochem. angiogenesis. Invest. Ophthalmol. Vis. Sci. 51, 1098–1105 (2010). Biophys. Res. Commun. 252,97–102 (1998). 11. Pietrzyk, J. J. et al. New insight into the pathogenesis of retinopathy of prematurity: 38. Wang, S. S. et al. Identification of novel methylation markers in cervical cancer assessment of whole-genome expression. Pediatr. Res. 73, 476–483 (2013). using restriction landmark genomic scanning. Cancer Res. 68, 2489–2497 (2008). 12. Maltby, S. et al. Opposing roles for CD34 in B16 melanoma tumor growth alter 39. Demokan, S. et al. Validation of nucleolar protein 4 as a novel methylated tumor early stage vasculature and late stage immune cell infiltration. PLoS ONE 6, suppressor gene in head and neck cancer. Oncol. Rep. 31, 1014–1020 (2014). e18160 (2011). 40. O'Reilly, M. A., Staversky, R. J., Stripp, B. R. & Finkelstein, J. N. Exposure to 13. Siemerink, M. J. et al. CD34 promotes pathological epi-retinal neovascularization hyperoxia induces p53 expression in mouse lung epithelium. Am. J. Respir. Cell in a mouse model of oxygen-induced retinopathy. PLoS ONE 11, e0157902 Mol. Biol. 18,43–50 (1998). (2016). 14. Rocha, S. F. et al. Esm1 modulates endothelial tip cell behavior and vascular permeability by enhancing VEGF bioavailability. Circ. Res. 115, 581–590 (2014). Open Access This article is licensed under a Creative Commons 15. Sorrentino, F. S. et al. Diabetic retinopathy and endothelin system: microangio- Attribution 4.0 International License, which permits use, sharing, pathy versus endothelial dysfunction. Eye (Lond.). 32, 1157–1163 (2018). adaptation, distribution and reproduction in any medium or format, as long as you give 16. Rattner, A. et al. Endothelin-2 signaling in the neural retina promotes the appropriate credit to the original author(s) and the source, provide a link to the Creative endothelial tip cell state and inhibits angiogenesis. Proc. Natl Acad. Sci. USA. 110, Commons license, and indicate if changes were made. The images or other third party E3830–E3839 (2013). material in this article are included in the article’s Creative Commons license, unless 17. Patel, C. et al. Activation of the endothelin system mediates pathological indicated otherwise in a credit line to the material. If material is not included in the angiogenesis during ischemic retinopathy. Am. J. Pathol. 184, 3040–3051 (2014). article’s Creative Commons license and your intended use is not permitted by statutory 18. Blom, J. J. et al. Inhibition of the adrenomedullin/nitric oxide signaling pathway in regulation or exceeds the permitted use, you will need to obtain permission directly early diabetic retinopathy. J. Ocul. Biol. Dis. Info. 4,70–82 (2011). from the copyright holder. To view a copy of this license, visit http://creativecommons. 19. Rungger-Brändle, E., Dosso, A. A. & Leuenberger, P. M. Glial reactivity, an early org/licenses/by/4.0/. feature of diabetic retinopathy. Invest. Ophthalmol. Vis. Sci. 41, 1971–1980 (2000). 20. Kasai, A. et al. Apelin is a crucial factor for hypoxia-induced retinal angiogenesis. Arterioscler. Thromb. Vasc. Biol. 30, 2182–2187 (2010). © The Author(s) 2019

Pediatric Research (2020) 87:485 – 493